It's amazing that the move of RISC-V Foundation to Switzerland (as RISC-V International) was actually called for.
Even more amazing is US lawmakers continuing to try & shoot the US in the foot.
RISC-V is now globally developed, that ship has sailed. Same for products based on it. The US can't prevent its 'proliferation'. Only reduce the degree to which US entities are part of, and profit from that.
Are US lawmakers that dumb? (ok I guess that's a rhetorical question).
I posted the numbers as a top level comment here [0], but the short version is that a lot of the signatories of this letter received (small) donations from Intel in the last election cycle.
My best guess is that Intel is actively lobbying against RISC-V using China as the boogeyman, and that these lawmakers were persuaded (sincerely or otherwise) to treat RISC-V as a threat.
I think the donations part of it isn’t even that important.
If you think about how legislation does (and can) work, lawmakers, even if they were the most gifted people on the planet, obviously are not experts on everything. As a result they have to rely on others.
But who should those others be? When it comes to something like chip manufacturing, you’re down to essentially 1 major U.S. company in this area that can provide guidance.
One could argue they could go to academics, but a single academic is good for going deep into a specific area and not so good at cross concerns.
And bringing together all those people is effort.
On the other hand you have Intel that’s got scores of researchers, marketing experts, and lobbyists who are basically giving you high quality data and information for free.
Even without money and therefore outright corruption playing any role, companies like these have an outsized influence on lawmakers for absolutely practical reasons.
Yes, while political donations in general are a concern to me, I think in this case the donations themselves are fairly innocuous. What the donations tell me is not that Intel bribed these representatives, it's that Intel identified these representatives as ones that would be particularly pliable to their legal persuasion efforts.
It's true that Intel has a lot of natural strengths for its lobbying efforts, but some representatives are going to be easier to persuade than others.
I think you'd be pressed to find a representative who didn't receive donations from Intel. Why would they gamble and guess on which ones will win next election or which ones will help them push their agenda?
RISC-V is much more of a threat to ARM microcontroller vendors (STMicro, NXP, Atmel et. al.) than it is to Intel. There is almost zero overlap in products/markets between RISC-V and x86. Even as you go up the stack, vendors like ARM International would care first (as it impacts mobile SOC licenses). Frankly even Apple probably cares more than Intel does.
Apple just switched to ARM away from Intel, and I'm sure Intel is very concerned others will follow their lead. If ARM is at risk from RISC-V, then Intel is as well, transitively.
That there is zero overlap now does not mean it will stay that way, as Apple demonstrated so well.
Yeah, but there's no product. There's no RISC-V CPU targetted at laptop or datacenter users that Intel would be worried about, but there absolutely are from ARM vendors.
So... how would hurting RISC-V, which helps ARM, be good for Intel? It's not. Pick a different boogeyman basically, Intel isn't the one here.
But by that logic Intel should be lobbying the government to outlaw every CPU architecture, or every competing product, or... every product they might conceivably want to compete with. The same logic works for every product! Or every company! It's all specious.
In point of fact, and I'll just say it again: this is dumb. Intel isn't remotely the most likely candidate here (again, my read is that on balance they'd love to see ARM competitors get some pressure on the other side). Trying to claim otherwise is just engaging in platform flamery nonsense. The logic doesn't work.
There was no competitive ARM CPU targeting laptop or datacenter users either, until Apple up and made one.
For Intel to ignore RISC-V right now would be to completely neglect the lessons of Apple Silicon. They appear to have waited until there was an actual product taking their market share before taking ARM seriously. I expect them to learn from that and not wait for a product this time.
ARM in the datacentre was somewhat of a hot topic for a while before Apple released their laptop CPUs; Amazon launched the first Gravitron to GA in 2018. Ampere launched publicly available ARM server CPUs in 2020, around the same time as Apple. Alibaba Cloud launched their instances based on their own CPU in 2021, as did Oracle (based on Ampere's CPUs).
Given the years of engineering effort involved in getting any of these projects to release, this definitely looks like an industry shift, not something driven by Apple.
Perhaps Intel's thinking here is that competition with ARM will drive innovation and cost-downs in that space, which will bleed 'up' and make their own competition's (mostly ARM) products even more competitive.
Note that the shipping RISC-V SoC currently most suited to datacentres, the SG2042, is more powerful than the first Graviton deployed to AWS in November 2018 (five years ago) -- the C910 cores in the SG2042 are similar to the Arm A72 cores in the Graviton, but the SG2042 has 64 of them vs 16 in the Graviton.
Multiple companies have M1-class RISC-V in the pipeline.
Business advice that is tantamount to "Oh, you should just make a better product" is basically bad business advice. Of course they should, everyone should. But the problem is that if they tried, it (1) costs money and (2) leverages no existing advantages.
These low-margin companies live and die on compatibility and integration. The reason you pick a STM32-whatever as your choice for your new product is generally at least partially because your existing staff know the tooling and your existing production people know the board integration requirements.
Swap that for some new RISC-V gadget and, well, you might win! Or you might find your customers walking across the street to buy someone else's gadget, because yours has no advantages.
Ergo, if you're STMicro, RISC-V is a threat in a way that "Oh, you should just use it yourself" doesn't address.
Microcontrollers are, all things considered, not that much about the ISA itself, especially when we've gotten to 32-bit controllers big enough to program in C or higher-level languages.
People aren't reading the raw assembly and juggling registers themselves anymore. But they do care about the onboard peripherals, documentation, and tooling. The STM32 ecosystem is huge because they've nailed those things.
I could see a STM32Vxxx range, same basic design but drop a RISC-V core in, and new-dev projects will adopt it if it's got the same functionality bit is a nickel cheaper.
I can 100% guarantee to you that's not the case. Existing shipping products in this world are based on proprietary toolchains, HALs like CMSIS that are ARM-only, Eclipse-based IDEs that teams refuse to move away from, build tools wrapped around custom linker scripts for one architecture that no one understands anymore, ...
This isn't a FAANG office where everyone's an expert and people change tools weekly. Commercial firmware work for "boring" devices is extremely conservative, and there's a ton of money left to be made shipping junk that's merely "compatible".
That audience is sort of immaterial to the argument.
If you have a firm that's committed to a frozen image of the platform on a specific date, they'll stick with existing products or modest riffs on it. The same premise is why we'll still see Z80 and 8051 derivatives until long past the heat death of the universe.
This is more about why a firm like STMicro doesn't need to freak out too much about RISC-V. There's no reason they 1) can't continue to sell ARM alongside RISC-V for the more change-averse audience and 2) leverage their established reputation in onboard-peripherals, toolchain, and documentation as their competitive edge.
Done right, this can be a real "commoditize your compliment" play: they can focus less effort on the boring "core CPU" functionality of their MCUs and divert it for the parts of the product they can offer a stronger case for.
> That audience is sort of immaterial to the argument.
You lost me. "That audience" is probably 80% of ST's market. If you're not talking about business as it exists then I don't know what your argument is about. It's certainly not responsive to the point I was making, which is that ST (and similar companies) is exposed to RISC-V as a business risk in a way that Intel is not.
> They will for sure should they not switch to the standard ISA.
OK, no, that's just not true. Google "adoption curve". Saying that technology X is likely to win out eventually is not the same thing as saying there's no revenue to be had selling "legacy" technology Y. Even today people still fix bugs in the occasional COBOL gadget.
And if you're one of the companies (STMicro is one) selling at least partially "legacy" technology (they'd say "mature"), then anything that accelerates motion along that curve is a threat.
Nah, they really are this uninformed for the most part (there are notable exceptions of course). Story time! I went to law school in DC. One of my classmates worked as a staffer in the office of a prominent (leadership-level) congressperson (she was from his district). She had also done an internship with the UN. This congressperson had made public statements about defunding the UN and blah blah blah. He took her out to lunch one day and asked her to explain what the UN was and how it worked in the 30 minutes they had so he could field questions he was getting about his public statements.
You don't have to fall back to that. It isn't possible to know all the issues facing national attention. Someone puts in a good/scary story about something and it is really hard to see why that story is wrong. Often the story isn't even wrong, it is just not the complete picture and the evils of what the story misses out on are even worse. There are a bit over 500 people with power in DC, they cannot possibly debate all the possible issues that come up.
China learning from US contributions to Risc-v are not a good thing. However the obvious proposed solution wouldn't stop China from learning (it might delay them a few months, but they have plenty of smart engineers who can figure this out), and so will harm the US more.
I think this discounts the genuine political opinions that many elected officials in the US government have, especially about the new cold war with China.
My trust in US lawmakers practically no longer exists. Now whenever they make a questionable decision, the first thing I try to understand is who stands to make money off of it. If lawmakers have genuine policy opinions that's wonderful, but they have a big PR problem from decades of corruption to address.
It's so beyond dumb that they think this in any way is related to national security, this has the potential very literally create some huge troubles for the US in the future...
Even more amazing is US lawmakers continuing to try & shoot the US in the foot.
RISC-V is now globally developed, that ship has sailed. Same for products based on it. The US can't prevent its 'proliferation'. Only reduce the degree to which US entities are part of, and profit from that.
Are US lawmakers that dumb? (ok I guess that's a rhetorical question).